Suggestion: estimated emerge times?

[Imek]

Hi,

This may seem like a fatally flawed or/and n00bish suggestion, but I thought I'd suggest it anyway. When I first installed Gentoo, I was at a loss as to how long it would take to compile. I found a website which had a limited list of how long it would take to compile each program on a 1GHz machine, so you could multiply/divide it to meet your clock speed. Then I thought, why not put this into portage, so that you could tell it to estimate how long it would take to compile on your PC. Sure, it wouldn't be that accurate, but a rough estimation would really ease the installation for new people, no doubt. I don't think it would be that hard to collect the right data wouldn't be that hard, would it? Or am I wrong? Criticise as required.

Thanks,

- I
_________________
- I
(very old) "Current" desktop | Adopt an unanswered post today!

[NeddySeagoon]

Imek,

There are lots of things that contribute to compile times, not just CPU speed.

Including -pipe in CCFLAGS makes a huge difference, providing you have the memory to make use of it.
_________________
Regards,

NeddySeagoon

Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail.

[ectospasm]

It could work... but there are a lot of contingencies to consider, like USE settings, CFLAGS, memory (virtual and RAM), and network connection speed/reliability. There'd need to be a lot of work done on portage, and the algorithm to figure it out for a given package is non-trivial, or so it seems to me. Adding it up for a set of packages shouldn't be difficult. The main drawback I see is that it can't possibly be accurate, since different programs, all other things being equal, will by the nature of the beast have different compile times.

I would search bugzilla and see if anyone has posted such a feature request. If not, you could submit one. Good luck in having it implemented any time soon though. (-:
_________________
Join the adopt an unanswered post initiative today
Join the EFF!
Join the Drug Policy Alliance!

[jdhardy]

A better idea would be to automate the use of LFS's SBUs (Standard Binutils Units). Basically, take a package everyone has (like binutils), time the emerge, and then scale everything to that (Mozilla, for example, would be about 5-10 SBUs). The gentoo devs have to build most of the major packages anyway for the CDs, and all the users are constantly compiling; after each emerge, upload the time taken *in SBUs* to a central server (and average them out, say), and store the SBUs in the ebuild metadata (or fetch them).

It won't be perfect (nothing is), but it would be an estimate at least. I don't know if the binutils ebuild strips CFLAGS though (it probably does), so those would have an effect as well. Maybe a better 'base' package is needed (Bash?).

[G-Style]

It could work, but you would have to base the times on a certain CPU speed, and many people here have many different CPU speeds to begin with. Which would make the estimated time not all that accurant.
_________________
Mastering Windows isn't impressive. But mastering Linux is.

[rfujimoto]

Hardware, USE flags, CFLAGS, dependencies. I disagree, there's too many uncontrollable variables. In my opinion it's better to not have something then to have something that can be consistently wrong.

There was at one time a patch for emerge to use a progress bar. It was based on a time per file algo I believe. My memory is a bit fuzzy though, so I could be wrong how it works. I'd run a search on "progress and bar" in the Doc Tips & Tricks forum.

EDIT: I found the thread

[pjp]

https://forums.gentoo.org/viewtopic.php?t=7159 (Thats topic 7,159 of 151,384)
_________________
Quis separabit? Quo animo?